Embed Notice

I have a friend who works in a US state government role. While auditing service provider records for medical and related services to citizens, she and some of her colleagues just saw potential evidence of AI being used to summarize client reports from a state database. The database is highly sensitive, covered by HIPAA and other laws, and contains some very personal info like (I think this is right) medical & psychiatric diagnoses, interactions with law enforcement, medical procedures, psychological treatment, etc.

My guess: someone is dumping protected information into chatGPT to get summaries so they don't have to do the work. Meanwhile, they just gave vulnerable people's data to openAI (or Google or Microsoft et al.). And they are apparently submitting these summaries as "hard facts" data, despite errors in the summaries, so the individuals being served (or not) might get flagged for the wrong things, put in the wrong treatment categories, denied services, etc.

Probably not an AI scraper let loose on the database, but insidious and awful in plenty of other ways.

#ai #llm #government #clusterfuck